Neuroblastoma (NB) is an extracranial solid tumor of early childhood that originates from the developing sympathetic nervous system. Activation of telomere maintenance mechanisms characterizes high-risk NB,1 and additional alterations in the p53 or RAS pathway, known to play a role in NB pathogenesis,2–5 further worsen disease outcome. Amplification of the MYCN gene, which encodes for the oncogenic transcription factor MYCN (also known as N-Myc), was the first genomic aberration found to be associated with poor prognosis.6 Despite these molecular insights, high-risk NB remains a major challenge in pediatric oncology. Serious long-term side effects caused by intensive radiochemotherapy emphasize the need for novel treatment approaches with reduced toxicity.7 8

Monoclonal antibodies directed against programmed cell death 1 (PD-1), programmed cell death 1 ligand 1 (PD-L1) or cytotoxic T-lymphocyte associated protein 4 (CTLA-4), also known as immune checkpoint inhibitors, have significantly improved the therapy of various cancers. Immune checkpoint inhibitors reinvigorate anti-tumor immunity depending on various parameters including tumor immunogenicity and the presence of tumor-infiltrating T cells.9 The latter is also known as a T-cell-inflamed or ‘hot’ tumor microenvironment (TME). A hot TME is further characterized by high interferon (IFN) pathway activity,10 a signaling cascade that plays a central role in the coordination of immune responses.11 In contrast, a T-cell infiltration-poor, ‘cold’ TME is usually associated with poor responses to immune checkpoint blockade (ICB) therapy. Recently, we and others showed that MYCN-amplified NBs are characterized by a ‘cold’ TME. Mechanistically, MYCN suppresses IFN activity and chemokine expression in NB cells, thus providing a possible explanation for this association,12–16 in line with the emerging notion that ICB therapy seems to be ineffective in patients with high-risk NB, although clinical data remain limited.17 18

We previously proposed that STING agonists could be promising drugs to induce conversion of ‘cold’ to ‘hot’ TMEs in NB.12 The cGAS/STING pathway plays a central role in the sensing of cytosolic DNA and its activation triggers a strong type I IFN response.19 Recently, Wang-Bishop et al20 demonstrated that intratumoral application of STING-activating nanoparticles induced immunogenic cell death, chemokine release and a T-cell-inflamed ‘hot’ TME in transplantable mouse models of NB improving ICB therapy. Depending on the clinical context, however, intratumoral injections of STING agonists could be challenging. Advanced strategies of tumor-targeted drug delivery could overcome this issue,21 but also alternative treatment strategies are needed.

Here, we tested the concept that epigenetic modifiers could be attractive drug targets to rewire NB immune phenotypes. By exploring NB transcriptome data, we found that euchromatic histone lysine methyltransferases EHMT2 and EHMT1 (also known as G9a and GLP) were negatively correlated with the expression of Th1-type chemokines (CXCL9, CXCL10), key drivers of a T-cell-inflamed ‘hot’ TME. EHMT expression was higher in MYCN-amplified human NBs and pharmacological inhibition of EHMTs enhanced IFN-γ-induced chemokine expression. In MYCN-amplified NB cells, robust IFN-γ responses required co-inhibition of EZH2 (enhancer of zeste homologue 2), a histone methyltransferase cooperating with EHMTs in gene repression. Finally, gene signatures from inhibitor-treated NB cells served as surrogate measures of high EHMT and EZH2 activity in patient samples from high-risk NBs. These gene signatures showed a strong association with MYCN amplification and a T-cell-poor, cold TME delineating tailored strategies for epigenetic immunomodulation of high-risk NBs.

Previously, we found that depletion of MYCN by RNA interference (RNAi) in NB cell lines enhanced IFN pathway activity, both at baseline and in response to IFN-γ or a synthetic STING agonist.12 Now, we asked whether we could identify pharmacological strategies to achieve the same effect. We hypothesized that epigenetic modifiers could be downstream effectors of MYCN and attractive drug targets to restore IFN responsiveness and chemokine expression in NBs.30 We therefore analyzed transcriptome data of a well-characterized cohort of INSS stage 4 NBs (metastatic disease) and we ranked genes based on their correlation with the IFN-γ-responsive chemokine CXCL10.31 Of note, one of the most negatively correlated gene encoded for the euchromatic histone lysine methyltransferase 2 (EHMT2), also known as G9a (figure 1A and online supplemental figure 1A,B). EHMT1 (GLP), another histone methyltransferase and closely related to EHMT2, was also among the negatively correlated genes with CXCL10. EHMT2 and EHMT1 are known to form a complex (G9a/GLP complex) involved in transcriptional silencing by introducing repressive H3K9me2 histone marks.32 33 Both EHMT2 and EHMT1 mRNA levels were positively correlated with MYCN expression (figure 1B) and significantly associated with MYCN amplification, in particular EHMT2 (figure 1C). Transcriptome profiling of our small NB cell line panel (n=8) confirmed higher EHMT2 expression in MYCN-amplified cells (figure 1D). Differences in expression of MYCN and MYC in our NB cell line panel were confirmed by western blot in line with literature (online supplemental figure 1C).28 Depletion of MYCN by RNAi moderately reduced the levels of EHMT2 and EZH2 but also EHMT1 and EZH1 (figure 1E–G). The extent of EHMT2 regulation by MYCN was in line with a recent study.34 EZH2, a H3K27me3 histone methyltransferase, was previously shown to be a MYCN target conferring epigenetic dependency.35 Analyzing publicly available MYCN ChIP-seq data,28 we found a MYCN-binding peak in the region of the transcriptional start site of EHMT2 and confirmed MYCN binding to the EZH2 promoter as reported (figure 1H,I).35 In MYCN-non-amplified cells with high MYC expression, MYC peaks in these regions were found variable (online supplemental figure 2A–F). No prominent peaks were detected for MYCN or MYC within the EHMT1 genomic region (online supplemental figure 2G,H). Together, the results suggested that MYCN, and likely also MYC if abundantly expressed, directly contribute to the expression of both EHMT2 and EZH2. Having addressed the regulation, we next interrogated a cancer type-dependent context of EHMT expression. Analysis of the CCLE data set showed that EHMT2 and EHMT1 mRNA expression was significantly higher in NB cells when compared with other cell lines from various types of cancers, which indicated potential lineage-specific functions of EHMTs (figure 1J,K).29 Altogether, our correlation analyses suggested a potential role for EHMT2 and EHMT1 in the repression of IFN-γ-responsive chemokines like CXCL10 in NB cells, whereby MYCN and possibly also high levels of MYC directly promote EHMT2 expression together with EZH2.

To test this hypothesis, we treated four human (SK-N-BE, IMR-32, NMB, SH-SY5Y) and two mouse NB cell lines derived from different MYCN-driven genetic models (mNB-A1, NHO2A)22 23 with UNC-0638 (UNC) or BIX-01294 (BIX), two selective and well-characterized EHMT inhibitors.36 37 As an on-target readout, we confirmed a reduction of H3K9me2 levels on drug exposure by western blot, although variability was noted across the cell lines tested (figure 2A). Both UNC and BIX enhanced IFN-γ-induced expression of CXCL10 mRNA and protein in MYCN-amplified SK-N-BE cells (figure 2B,C). We also confirmed this result in an independent MYCN-amplified human NB cell line IMR-32 (figure 2D,E) and the mouse MYCN-transgenic NB cell line mNB-A1 (figure 2F,G). In summary, these results supported the idea that EHMTs are involved in the repression of IFN-γ-responsive chemokines associated with high MYCN level in human and mouse NB cells. We also noted that EHMT inhibition enhanced IFN-γ-induced expression of CXCL10 mRNA and protein in MYCN non-amplified SH-SY5Y and SK-N-AS cells, both expressing high levels of MYC (figure 2H–K and online supplemental figure 1C). EHMT inhibitor alone induced CXCL10 mRNA expression in SK-N-AS cells (figure 2J).

Next, we asked whether EHMT inhibitors also affected the proliferation of NB cells. To this end, we analyzed our panel of four MYCN-amplified (SK-N-BE, IMR-32, NMB, CHP-134) and four MYCN-non-amplified human NB cell lines (GI-ME-N, SK-N-AS, SK-N-FI, SH-SY5Y) that were exposed to increasing concentrations of UNC or BIX (figure 3A,B). We found that EHMT inhibitors strongly impaired the proliferation of MYCN-amplified NB cells, but had a moderate effect on the proliferation of the MYCN-non-amplified NB cell lines tested besides SH-SY5Y cells (figure 3C,D). Thus, proliferation of MYCN-amplified NB cells was dependent on the activity of EHMTs as also reported by others very recently.34 This prompted us to ask whether EHMT inhibition also had concordant effects on IFN-γ responsiveness. We treated each cell line with IFN-γ or IFN-γ plus UNC in biological replicates and analyzed transcriptome changes by 3′mRNA-seq. Interestingly, EHMT inhibition enhanced IFN-γ-induced expression of CXCL10 and interferon responsive genes (MSigDb Hallmark IFN-γ response) to a larger extent in MYCN-non-amplified than in MYCN-amplified NB cell lines (figure 3E,F). Actually, global IFN-γ responsiveness on EHMT inhibition of MYCN-amplified NB cell lines was rather moderate. In order to analyze the contribution of other oncogenic pathways involved in epigenetic gene silencing, we compared the transcriptomes of the NB cells treated with IFN-γ plus UNC by GSEA using the MSigDb C6 oncogenic signatures collection. Notably, among the significantly (FDR<0.05) regulated gene sets, PRC2_EZH2-UP.V1_DN was enriched in MYCN-amplified versus MYCN-non-amplified NB cells, which indicated higher EZH2 activity in the former (figure 3G,H). Together with embryonic ectoderm development protein (EED) and SUZ12 polycomb repressive complex 2 (PRC2) subunit (SUZ12), EZH2 (as well as EZH1) is a core component of PRC2 involved in gene silencing. Of note, PRC2 physically interacts with the EHMT (G9a/GLP) complex and functionally cooperates in transcriptional repression.33 Thus, GSEA suggested to combine PRC2 inhibitors (eg, EZH2 inhibitors) with EHMT inhibitors in order to achieve robust transcriptional responses to IFN-γ also in MYCN-amplified NB cells.

To test this idea, we treated SK-N-BE and IMR-32 cells with the EED inhibitor EED226 (EED) or the EZH inhibitors GSK-503 (GSK), CPI-1205 (CPI) or EPZ011989 (EPZ) and confirmed on-target reduction of H3K27me3 levels by western blot analysis (figure 4A). All three EZH inhibitors are more effective against EZH2 than EZH1.36 To test combined effects, we used UNC-0638 (UNC) and EPZ011989 (EPZ). At least in the time frame analyzed, UNC reduced H3K9me2 but not H3K27me3 levels, and vice versa EPZ, confirming on-target activity of the inhibitors (figure 4B). NB cells treated with a combination of both drugs (UNC+EPZ) consistently showed reduced levels of both histone methylation marks. In combination with EHMT inhibition by UNC, all EZH or EED inhibitors enhanced IFN-γ-induced CXCL10 mRNA expression (figure 4C). Next, we performed 3′mRNA-seq transcriptome analysis of SK-N-BE and IMR-32 cells in the absence or presence of IFN-γ, in addition to treatment with vehicle, UNC, EPZ or both drugs (UNC+EPZ). Single drug treatment with either UNC or EPZ enhanced the transcriptional response to IFN-γ, but the combination treatment UNC+EPZ was more effective in both NB cell lines (figure 4D,E). Effects were not restricted to IFN-γ-inducible Th1-type chemokines CXCL9, CXCL10 or CXCL11, but included typical IFN-γ-inducible genes from various functional categories such as transcription factors (eg, STAT1, IRF1), antigen presentation (B2M, NLRC5, PSMB9, HLA-B), pattern recognition receptors (DDX58 encoding RIG-I) and immune checkpoints (CD274 encoding PD-L1). The combination treatment UNC+EPZ also strongly enhanced IFN-γ-induced CXCL10 release into cell culture supernatants assessed by ELISA (figure 4F,G).

Next, we addressed the epigenetic context how UNC+EPZ promotes IFN-γ responses. In particular, we were interested in the CXCL9, CXCL10 and CXCL11 Th1-type chemokines clustered on chromosome 4. We therefore analyzed published histone ChIP-seq data that provided epigenomic profiling of several NB cell lines, both MYCN-amplified and MYCN-non-amplified.28 The data set included ChIP-seq for the repressive histone mark H3K27me3 as well as for activating histone marks (H3K27Ac, H3K4me3), whereas H3K9me2 was not assessed in this study. We found enriched H3K27me3 signals in the entire CXCL9, CXCL10 and CXCL11 genomic region in five out of six MYCN-amplified NB cell lines (SK-N-BE, NB1643, COG-N415, NGP, Kelly and LAN-5) (figure 5A,B and online supplemental figure 3A–D), but only in one out of four MYCN-non-amplified NB cell lines (NB-69, NBLS, SK-FI, SK-N-AS) (online supplemental figure 3E-H). H3K27me3 signals were low in the region of the neighboring gene SDAD1, which in contrast showed prominent ChIP-seq peaks of the activating histone marks H3K27ac and H3K4me3 in the promotor region (figure 5A,B and online supplemental figure 3A–D). Further, SDAD1 showed a prominent MYCN-binding beak in its promoter region (figure 5C), and previous transcriptome profiling of MYCN-depleted NB cells by RNAi suggested that SDAD1 is a potential MYCN target gene (figure 5D). Importantly, no MYCN-binding or MYC-binding peaks were found in the genomic region of CXCL9, CXCL10 and CXCL11 (figure 5D and online supplemental figure 3I). Recently, MYC was found to directly interact with EHMTs mediating gene repression and promoting tumorigenesis.38 At least in the case of regulation of CXCL9, CXCL10 and CXCL11 expression in NB cells, neither MYCN nor MYC seemed to be directly involved in the recruitment of the EHMT/EZH2 repressive complexes, but indirectly by increasing the levels of EHMT2 and EZH2 in NB cells. As we had also profiled IFN-γ responses of two MYCN-non-amplified NB cell lines SK-N-FI and SK-N-AS (high vs low H3K27me3 signals at CXCL9, CXCL10 and CXCL11 loci) overlapping with the ChIP-seq study,28 we consistently noticed that low H3K27me3 signals in SK-N-AS cells correlated with strong CXCL9, CXCL10 and CXCL11 gene induction by IFN-γ treatment alone (online supplemental figure 3G,H,J).

Given that neither H3K9me2 was profiled in Upton et al.28 nor exposures to EZH2 or EHMT inhibitors were tested, we performed ChIP-qPCRs for H3K27me3 and H3K9me2 assessing regulatory regions within CXCL9, CXCL10 and SDAD1 as control at four different genomic sites each (figure 5E). Control ChIP-qPCRs for histone H3 are shown in online supplemental figure 4. SK-N-BE cells were treated with UNC, EPZ or UNC+EPZ for 6 days in biological triplicates and chromatin was harvested for ChIP-qPCR analyses. First, H3K27me3 level (% input) was higher at all regions in CXCL9 and CXCL10 compared with SDAD1 in untreated cells (figure 5F). Exposure to EPZ or UNC+EPZ resulted in a strong reduction of H3K27me3 level at all loci reaching a similar baseline level (figure 5F). Consistently, H3K9me2 level (%input) was also higher at all tested regions in CXCL9 and CXCL10 compared with the SDAD1 control regions in untreated cells (figure 5G). Notably, UNC or UNC+EPZ treatment resulted in a reduction of H3K9me2 level in all tested regions in CXCL9 and CXCL10, but regions in SDAD1 remained unaltered (figure 5G). Furthermore, H3K9me2 and H3K27me3 level closely correlated at regions in CXCL9 and CXCL10 underscoring the importance of both repressive histone marks in the regulation of CXCL9 and CXCL10 chemokine gene expression (figure 5H,I).

Next, we aimed to establish UNC and EPZ drug response gene signatures from NB cell lines as surrogate measures of EHMT and EZH2 activity for the analysis of transcriptomes from NB patient samples (figure 6A). We reasoned that expression level of EHMTs or EZH2 might or might not correlate with EHMT and EZH2 activity in NB patient samples and drug response signatures could be better estimates of activity. We also took into consideration that EHMT and EZH2 inhibitors have both direct and indirect effects on transcriptional responses, for which reason we included both upregulated and downregulated genes into our analysis.27 Thus, we treated SK-N-BE and IMR-32 NB cells with UNC, EPZ or both drugs and performed 3′mRNA-seq as described above, without IFN-γ exposure. We then determined the differentially expressed genes (DEGs) that were either upregulated or downregulated (FDR<0.05) by UNC, EPZ or the combination drug exposure UNC+EPZ in SK-N-BE and IMR-32 cells (online supplemental tables S1–S3). Consequently, we obtained six drug response signatures, for example, ‘Up_by_UNC-0638’ and ‘Down_by_UNC-0638’, based on the overlap of the DEGs (figure 6B–D and online supplemental tables S4–S6). Of note, MYCN was 1 of the 47 genes in the ‘Down_by_UNC-0638’ signature. However, MYCN was omitted from downstream analysis because MYCN expression is known to strongly stratify NB samples. In order to determine surrogate EHMT activity scores in transcriptomes from NB patient samples, we divided the averaged log2 expression values of ‘Down_by_UNC-0638’ genes (positive correlation with EHMT activity) by the averaged log2 expression values of ‘Up_by_UNC-0638’ genes (negative correlation with EHMT activity) for each NB sample (figure 6E). Thus, a higher score would indicate a higher EHMT activity. Heatmaps visualize opposing expression pattern of ‘Down_by_UNC-0638’ genes and ‘Up_by_UNC-0638’ genes in high-risk NBs ranked by increasing EHMT activity scores (figure 6F,G). Scores for surrogate EZH2 activity and combined EHMT+EZH2 activity were calculated in the same manner as EHMT activity scores. EPZ is rather selective for EZH2, and therefore, EZH activity scores were assumed to reflect EZH2 activity.39

In high-risk NB samples, EHMT2, EHMT1 and MYCN mRNA levels were positively correlated with the EHMT activity score with the highest values observed for EHMT2 and MYCN, respectively (figure 6H). Interestingly, EHMT activity score was negatively correlated with surrogate cytotoxic T-cell content (CD8A, PRF1, GZMA) and Th1-type chemokines (CXCL9, CXCL10) further supporting a link between high EHMT activity and a ‘cold’ immune phenotype in NB. We repeated the analysis for the EZH2 activity score showing a positive correlation with EZH2 gene expression, but a negative correlation with EZH1 gene expression (figure 7I). T-cell content (CD8A, PRF1, GZMA) and Th1-type chemokines (CXCL9, CXCL10) were also negatively correlated with the EZH2 activity score, but weaker than the EHMT activity score. Finally, we repeated the analyses for the combined EHMT+EZH2 activity score, which revealed a positive correlation with MYCN, EHMT2, EHMT1 and EZH2 gene expression (figure 7J). In comparison with individual EHMT or EZH activity scores, the combined EHMT+EZH2 activity score showed the strongest negative correlation with immune contexture marker genes for cytotoxic T cells (CD8A, PRF1, GZMA) and Th1-type chemokines (CXCL9, CXCL10). Of interest, the combined EHMT+EZH2 activity score also showed significant negative correlations with the immune marker genes when restricting the analyses to MYCN-amplified cases only (online supplemental figure 5). Together with our experimental data, the bioinformatic analyses suggested that high EHMT and EZH2 activity contribute to the T-cell poor cold immune phenotype of high-risk NBs with MYCN amplification providing a rationale for targeted epigenetic immunomodulation.

To mimic this scenario, we devised a functional assay co-culturing a bulk population of UNC+EPZ or vehicle-treated SK-N-BE or SH-SY5Y cells with a low frequency of human T cells as a natural source of IFN-γ rather than adding recombinant IFN-γ (figure 7A). MYCN-non-amplified SH-SY5Y cells expressed high levels of MYC (online supplemental figure 1C) and showed the strongest antiproliferative response to EHMT inhibitors among the MYCN-non-amplified NB cell lines tested (figure 3A–D), for which reason SH-SY5Y cells were also a valuable model. We decided to compare non-activated with activated T cells (anti-CD3, anti-CD28) in co-cultures, as IFN-γ production would be only detected in the latter. Indeed, flow cytometric analyses confirmed that CD69⁺ IFN-γ⁺ CD4⁺ or CD8⁺ T cells were exclusively detected when T cells were exposed to anti-CD3 and anti-CD28 (figure 7B). On average, about 10% of T cells (CD4⁺ and CD8⁺) were CD69⁺IFN-γ⁺ in the pool of anti-CD3 and anti-CD28-activated T cells (figure 7C). We then decided to co-culture SK-N-BE or SH-SY5Y cells with non-activated or activated T cells at a ratio of 10:1. By this, we obtained a final ratio of SK-N-BE or SH-SY5Y cells to CD69⁺ IFN-γ⁺ T cells of about 100:1 when adding the activated T-cell pool. After co-culture, we analyzed CXCL9 and CXCL10 mRNA expression in NB cells showing the strongest induction when SK-N-BE or SH-SY5Y cells were pretreated with UNC+EPZ and co-cultured with activated T cells (figure 7D–G). We also analyzed MHC class I surface expression on SK-N-BE or SH-SY5Y showing again the highest levels when the NB cells were pretreated with UNC+EPZ and co-cultured with activated T cells (figure 7H,I). Thus, a low frequency of CD69⁺IFN-γ⁺ T cells (~1%) strongly promoted CXCL9 and CXCL10 chemokine expression and MHC I expression in co-culture assays if the NB cells were pretreated with EHMT and EZH inhibitors. Based on this, we propose that EHMT and EZH inhibitors could instigate a feedforward loop of IFN-γ⁺ and Th1-type chemokine expression in the TME, which would be therapeutically most beneficial in the case of NBs with high EHMT and EZH activity such as MYCN-amplified NBs (figure 7J). Nevertheless, also a subset of MYCN-non-amplified NBs might benefit from this strategy, but the precise context of high EHMT and EZH2 activity within this subset remains to be determined.

MYCN amplification has been previously shown by us and others to be associated with a T-cell poor TME phenotype in metastatic NB.12 13 40 As drugging MYCN remains a challenge, our study asked whether epigenetic modifiers could be promising drug targets for immunomodulation of NB. Through the exploration of NB transcriptomes, we identified euchromatic histone lysine methyltransferases EHMT2 and EHMT1 as potential suppressors of IFN-γ responsive Th1-type chemokines like CXCL10, which was confirmed experimentally using selective small molecule inhibitors. We also provided evidence for a functional cooperation between EHMTs and EZH2 in the suppression of IFN pathway activity, in line with a report showing that these epigenetic modifiers mediate gene silencing in a concerted manner.33 Levels of the repressive histone marks H3K9me2 and H3K27me3 were enriched at the CXCL9, CXCL10 and CXCL11 chemokine cluster in SK-N-BE cells when compared with neighboring regions and correlated. While our manuscript was in revision, two preprints also propose epigenetic mechanisms, EZH2 and phenotypic cell states, as regulators and determinants of NB immune signaling.41 42 Together with our work, this underscores the emerging concept of using epigenetic drugs for targeted immunomodulation of high-risk NBs.

The first functional associations between EHMT activity and IFN signaling were reported in the context of antiviral responses. An early study described that EHMT2 was responsible for silencing of IFN-β (IFNB1) gene transcription through a cooperation with the transcription factor PRDM1 (BLIMP-1).43 More recent studies have established a role for H3K9me2 in the repression of IFNs and IFN-stimulated genes, which could be overcome by genetic or pharmacological inhibition of EHMT2, for example, in fibroblasts.44 With regard to cancer therapy, EHMT inhibition by BIX-01294 has been shown to sensitize chronic myeloid leukemia cells to type I IFNs.45 A dual inhibitor (CM-272) of EHMTs and DNA methyltransferases (DNMTs) was demonstrated to be active against multiple preclinical models of hematological neoplasia as well as in enhancing IFN responsiveness.46 Additionally, CM-272 also promoted TME phenotype conversion from ‘cold’ to ‘hot’ in a mouse model of bladder cancer synergizing with anti-PD-1 ICB therapy.47 Despite these published links between EHMTs and suppression of IFN responses, such a role is yet to be established for NB. Importantly, our study has defined the clinical context in which EHMT2, EHMT1 together with EZH2 are expressed at high levels, mostly MYCN-amplified high-risk NBs with a T-cell-poor and non-inflamed TME phenotype. Our low-frequency T-cell co-culture assays depicted how EHMT and EZH2 inhibition could amplify weak IFN feedforward signaling in the TME of high-risk NBs. We believe that this epigenetic treatment approach could have important implications for the future development of immune checkpoint inhibitor therapy in NB to overcome the current limitations.17 18 Our limited data on NB cell lines also suggest that EHMT inhibition alone might be sufficient for MYCN-low and MYC-low NBs, but this would require future in-depth studies.

Although antiproliferative and cytotoxic effects of EHMT inhibitors in MYCN-amplified NB cells were reported very recently,34 our EHMT activity score further substantiated the notion that EHMTs are effectors of the MYCN-driven malignant phenotype in NB. Hence, exploiting this genotype-dependent vulnerability could be a promising therapeutic strategy for MYCN-amplified NBs. Even though our re-analysis of public ChIP-seq data indicated binding of MYC to the regulatory regions of EHMT2 and EZH2, less total MYC versus MYCN and/or functional differences between MYC and MYCN may explain higher expression and dependence on EHMT2 and EZH2 activity in MYCN-amplified NB cell lines. A CRISPR-Cas9 screen recently showed that MYCN-amplified NB cells were also highly dependent on EZH2 activity,35 a finding also supported by our observations whereby the combined EHMT+EZH2 activity score strongly correlated with EHMTs, EZH2 and MYCN expression. MYCN has been also shown to physically interact with EZH2-containing PRC2.48 Therefore, our work and that of others encourage further exploration of epigenetic drugs for the treatment of NB.34 35 49–51

In summary, our findings establish a link between increased EHMT and EZH2 activity and a T-cell-poor, cold TME immune phenotype in high-risk NB, by focusing on the epigenetic regulation of Th1-type chemokine genes with CXCL9 and CXCL10 being two key drivers of T-cell recruitment.52 Our experimental data further suggests that EHMTs and EZH2 are causally involved in the repression of interferon signaling, which promotes the establishment of a non-inflamed ‘cold’ TME by interfering with IFN-γ feedforward signaling. Therefore, our study provides a scientific basis to explore EHMT inhibitors, alone (maybe in a subset of MYCN-non-amplified NBs) or in combination with EZH2 inhibitors (in particular MYCN-amplified NBs), for targeted immunomodulation of NBs to induce a ‘hot’ TME phenotype for improved responsiveness to ICB therapy. Our proposed epigenetic treatment strategy also has the potential to synergize with efforts employing STING agonists or oncolytic viruses for multimodal immunotherapy of NB.20 53 Nevertheless, further preclinical work will be needed to delineate optimized drug combinations and regimens that could achieve long-term immunological control of high-risk NB.